diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwen_utils.py b/src/diffusers/pipelines/qwenimage/pipeline_qwen_utils.py index a3eccc4f65..81da3d85a9 100644 --- a/src/diffusers/pipelines/qwenimage/pipeline_qwen_utils.py +++ b/src/diffusers/pipelines/qwenimage/pipeline_qwen_utils.py @@ -12,6 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. +import inspect import math from typing import List, Optional, Union @@ -20,6 +21,104 @@ import torch from ...utils import deprecate +# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift +def calculate_shift( + image_seq_len, + base_seq_len: int = 256, + max_seq_len: int = 4096, + base_shift: float = 0.5, + max_shift: float = 1.15, +): + m = (max_shift - base_shift) / (max_seq_len - base_seq_len) + b = base_shift - m * base_seq_len + mu = image_seq_len * m + b + return mu + + +# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps +def retrieve_timesteps( + scheduler, + num_inference_steps: Optional[int] = None, + device: Optional[Union[str, torch.device]] = None, + timesteps: Optional[List[int]] = None, + sigmas: Optional[List[float]] = None, + **kwargs, +): + r""" + Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles + custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. + + Args: + scheduler (`SchedulerMixin`): + The scheduler to get timesteps from. + num_inference_steps (`int`): + The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps` + must be `None`. + device (`str` or `torch.device`, *optional*): + The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. + timesteps (`List[int]`, *optional*): + Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed, + `num_inference_steps` and `sigmas` must be `None`. + sigmas (`List[float]`, *optional*): + Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed, + `num_inference_steps` and `timesteps` must be `None`. + + Returns: + `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the + second element is the number of inference steps. + """ + if timesteps is not None and sigmas is not None: + raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values") + if timesteps is not None: + accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) + if not accepts_timesteps: + raise ValueError( + f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" + f" timestep schedules. Please check whether you are using the correct scheduler." + ) + scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs) + timesteps = scheduler.timesteps + num_inference_steps = len(timesteps) + elif sigmas is not None: + accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) + if not accept_sigmas: + raise ValueError( + f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" + f" sigmas schedules. Please check whether you are using the correct scheduler." + ) + scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs) + timesteps = scheduler.timesteps + num_inference_steps = len(timesteps) + else: + scheduler.set_timesteps(num_inference_steps, device=device, **kwargs) + timesteps = scheduler.timesteps + return timesteps, num_inference_steps + + +# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents +def retrieve_latents( + encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample" +): + if hasattr(encoder_output, "latent_dist") and sample_mode == "sample": + return encoder_output.latent_dist.sample(generator) + elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax": + return encoder_output.latent_dist.mode() + elif hasattr(encoder_output, "latents"): + return encoder_output.latents + else: + raise AttributeError("Could not access latents of provided encoder_output") + + +def calculate_dimensions(target_area, ratio): + width = math.sqrt(target_area * ratio) + height = width / ratio + + width = round(width / 32) * 32 + height = round(height / 32) * 32 + + return width, height, None + + class QwenImageMixin: @property def guidance_scale(self): @@ -340,13 +439,3 @@ class QwenImageEditPlusPipelineMixin(QwenImageEditPipelineMixin): prompt_embeds = prompt_embeds.to(dtype=dtype, device=device) return prompt_embeds, encoder_attention_mask - - -def calculate_dimensions(target_area, ratio): - width = math.sqrt(target_area * ratio) - height = width / ratio - - width = round(width / 32) * 32 - height = round(height / 32) * 32 - - return width, height, None diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage.py index 6ac5acf364..822f1af69f 100644 --- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage.py +++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage.py @@ -12,7 +12,6 @@ # See the License for the specific language governing permissions and # limitations under the License. -import inspect from typing import Any, Callable, Dict, List, Optional, Union import numpy as np @@ -27,7 +26,7 @@ from ...utils import is_torch_xla_available, logging, replace_example_docstring from ...utils.torch_utils import randn_tensor from ..pipeline_utils import DiffusionPipeline from .pipeline_output import QwenImagePipelineOutput -from .pipeline_qwen_utils import QwenImagePipelineMixin +from .pipeline_qwen_utils import QwenImagePipelineMixin, calculate_shift, retrieve_timesteps if is_torch_xla_available(): @@ -57,80 +56,6 @@ EXAMPLE_DOC_STRING = """ """ -# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift -def calculate_shift( - image_seq_len, - base_seq_len: int = 256, - max_seq_len: int = 4096, - base_shift: float = 0.5, - max_shift: float = 1.15, -): - m = (max_shift - base_shift) / (max_seq_len - base_seq_len) - b = base_shift - m * base_seq_len - mu = image_seq_len * m + b - return mu - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps -def retrieve_timesteps( - scheduler, - num_inference_steps: Optional[int] = None, - device: Optional[Union[str, torch.device]] = None, - timesteps: Optional[List[int]] = None, - sigmas: Optional[List[float]] = None, - **kwargs, -): - r""" - Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles - custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. - - Args: - scheduler (`SchedulerMixin`): - The scheduler to get timesteps from. - num_inference_steps (`int`): - The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps` - must be `None`. - device (`str` or `torch.device`, *optional*): - The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. - timesteps (`List[int]`, *optional*): - Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed, - `num_inference_steps` and `sigmas` must be `None`. - sigmas (`List[float]`, *optional*): - Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed, - `num_inference_steps` and `timesteps` must be `None`. - - Returns: - `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the - second element is the number of inference steps. - """ - if timesteps is not None and sigmas is not None: - raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values") - if timesteps is not None: - accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accepts_timesteps: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" timestep schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - elif sigmas is not None: - accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accept_sigmas: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" sigmas schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - else: - scheduler.set_timesteps(num_inference_steps, device=device, **kwargs) - timesteps = scheduler.timesteps - return timesteps, num_inference_steps - - class QwenImagePipeline(DiffusionPipeline, QwenImagePipelineMixin, QwenImageLoraLoaderMixin): r""" The QwenImage pipeline for text-to-image generation. diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet.py index a68dca6407..cec81f72da 100644 --- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet.py +++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet.py @@ -12,7 +12,6 @@ # See the License for the specific language governing permissions and # limitations under the License. -import inspect from typing import Any, Callable, Dict, List, Optional, Union import numpy as np @@ -28,7 +27,7 @@ from ...utils import is_torch_xla_available, logging, replace_example_docstring from ...utils.torch_utils import randn_tensor from ..pipeline_utils import DiffusionPipeline from .pipeline_output import QwenImagePipelineOutput -from .pipeline_qwen_utils import QwenImagePipelineMixin +from .pipeline_qwen_utils import QwenImagePipelineMixin, calculate_shift, retrieve_latents, retrieve_timesteps if is_torch_xla_available(): @@ -102,94 +101,6 @@ EXAMPLE_DOC_STRING = """ """ -# Coped from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift -def calculate_shift( - image_seq_len, - base_seq_len: int = 256, - max_seq_len: int = 4096, - base_shift: float = 0.5, - max_shift: float = 1.15, -): - m = (max_shift - base_shift) / (max_seq_len - base_seq_len) - b = base_shift - m * base_seq_len - mu = image_seq_len * m + b - return mu - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents -def retrieve_latents( - encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample" -): - if hasattr(encoder_output, "latent_dist") and sample_mode == "sample": - return encoder_output.latent_dist.sample(generator) - elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax": - return encoder_output.latent_dist.mode() - elif hasattr(encoder_output, "latents"): - return encoder_output.latents - else: - raise AttributeError("Could not access latents of provided encoder_output") - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps -def retrieve_timesteps( - scheduler, - num_inference_steps: Optional[int] = None, - device: Optional[Union[str, torch.device]] = None, - timesteps: Optional[List[int]] = None, - sigmas: Optional[List[float]] = None, - **kwargs, -): - r""" - Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles - custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. - - Args: - scheduler (`SchedulerMixin`): - The scheduler to get timesteps from. - num_inference_steps (`int`): - The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps` - must be `None`. - device (`str` or `torch.device`, *optional*): - The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. - timesteps (`List[int]`, *optional*): - Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed, - `num_inference_steps` and `sigmas` must be `None`. - sigmas (`List[float]`, *optional*): - Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed, - `num_inference_steps` and `timesteps` must be `None`. - - Returns: - `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the - second element is the number of inference steps. - """ - if timesteps is not None and sigmas is not None: - raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values") - if timesteps is not None: - accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accepts_timesteps: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" timestep schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - elif sigmas is not None: - accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accept_sigmas: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" sigmas schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - else: - scheduler.set_timesteps(num_inference_steps, device=device, **kwargs) - timesteps = scheduler.timesteps - return timesteps, num_inference_steps - - class QwenImageControlNetPipeline(DiffusionPipeline, QwenImagePipelineMixin, QwenImageLoraLoaderMixin): r""" The QwenImage pipeline for text-to-image generation. diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet_inpaint.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet_inpaint.py index a02e9b18ec..6b06392fd6 100644 --- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet_inpaint.py +++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet_inpaint.py @@ -12,7 +12,6 @@ # See the License for the specific language governing permissions and # limitations under the License. -import inspect from typing import Any, Callable, Dict, List, Optional, Union import numpy as np @@ -28,7 +27,7 @@ from ...utils import is_torch_xla_available, logging, replace_example_docstring from ...utils.torch_utils import randn_tensor from ..pipeline_utils import DiffusionPipeline from .pipeline_output import QwenImagePipelineOutput -from .pipeline_qwen_utils import QwenImagePipelineMixin +from .pipeline_qwen_utils import QwenImagePipelineMixin, calculate_shift, retrieve_timesteps if is_torch_xla_available(): @@ -75,94 +74,6 @@ EXAMPLE_DOC_STRING = """ """ -# Coped from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift -def calculate_shift( - image_seq_len, - base_seq_len: int = 256, - max_seq_len: int = 4096, - base_shift: float = 0.5, - max_shift: float = 1.15, -): - m = (max_shift - base_shift) / (max_seq_len - base_seq_len) - b = base_shift - m * base_seq_len - mu = image_seq_len * m + b - return mu - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents -def retrieve_latents( - encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample" -): - if hasattr(encoder_output, "latent_dist") and sample_mode == "sample": - return encoder_output.latent_dist.sample(generator) - elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax": - return encoder_output.latent_dist.mode() - elif hasattr(encoder_output, "latents"): - return encoder_output.latents - else: - raise AttributeError("Could not access latents of provided encoder_output") - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps -def retrieve_timesteps( - scheduler, - num_inference_steps: Optional[int] = None, - device: Optional[Union[str, torch.device]] = None, - timesteps: Optional[List[int]] = None, - sigmas: Optional[List[float]] = None, - **kwargs, -): - r""" - Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles - custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. - - Args: - scheduler (`SchedulerMixin`): - The scheduler to get timesteps from. - num_inference_steps (`int`): - The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps` - must be `None`. - device (`str` or `torch.device`, *optional*): - The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. - timesteps (`List[int]`, *optional*): - Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed, - `num_inference_steps` and `sigmas` must be `None`. - sigmas (`List[float]`, *optional*): - Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed, - `num_inference_steps` and `timesteps` must be `None`. - - Returns: - `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the - second element is the number of inference steps. - """ - if timesteps is not None and sigmas is not None: - raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values") - if timesteps is not None: - accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accepts_timesteps: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" timestep schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - elif sigmas is not None: - accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accept_sigmas: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" sigmas schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - else: - scheduler.set_timesteps(num_inference_steps, device=device, **kwargs) - timesteps = scheduler.timesteps - return timesteps, num_inference_steps - - class QwenImageControlNetInpaintPipeline(DiffusionPipeline, QwenImagePipelineMixin, QwenImageLoraLoaderMixin): r""" The QwenImage pipeline for text-to-image generation. diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit.py index 7e3e3a1bb0..f246ed2773 100644 --- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit.py +++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit.py @@ -12,7 +12,6 @@ # See the License for the specific language governing permissions and # limitations under the License. -import inspect from typing import Any, Callable, Dict, List, Optional, Union import numpy as np @@ -27,7 +26,13 @@ from ...utils import is_torch_xla_available, logging, replace_example_docstring from ...utils.torch_utils import randn_tensor from ..pipeline_utils import DiffusionPipeline from .pipeline_output import QwenImagePipelineOutput -from .pipeline_qwen_utils import QwenImageEditPipelineMixin, calculate_dimensions +from .pipeline_qwen_utils import ( + QwenImageEditPipelineMixin, + calculate_dimensions, + calculate_shift, + retrieve_latents, + retrieve_timesteps, +) if is_torch_xla_available(): @@ -64,94 +69,6 @@ EXAMPLE_DOC_STRING = """ """ -# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift -def calculate_shift( - image_seq_len, - base_seq_len: int = 256, - max_seq_len: int = 4096, - base_shift: float = 0.5, - max_shift: float = 1.15, -): - m = (max_shift - base_shift) / (max_seq_len - base_seq_len) - b = base_shift - m * base_seq_len - mu = image_seq_len * m + b - return mu - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps -def retrieve_timesteps( - scheduler, - num_inference_steps: Optional[int] = None, - device: Optional[Union[str, torch.device]] = None, - timesteps: Optional[List[int]] = None, - sigmas: Optional[List[float]] = None, - **kwargs, -): - r""" - Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles - custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. - - Args: - scheduler (`SchedulerMixin`): - The scheduler to get timesteps from. - num_inference_steps (`int`): - The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps` - must be `None`. - device (`str` or `torch.device`, *optional*): - The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. - timesteps (`List[int]`, *optional*): - Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed, - `num_inference_steps` and `sigmas` must be `None`. - sigmas (`List[float]`, *optional*): - Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed, - `num_inference_steps` and `timesteps` must be `None`. - - Returns: - `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the - second element is the number of inference steps. - """ - if timesteps is not None and sigmas is not None: - raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values") - if timesteps is not None: - accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accepts_timesteps: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" timestep schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - elif sigmas is not None: - accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accept_sigmas: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" sigmas schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - else: - scheduler.set_timesteps(num_inference_steps, device=device, **kwargs) - timesteps = scheduler.timesteps - return timesteps, num_inference_steps - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents -def retrieve_latents( - encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample" -): - if hasattr(encoder_output, "latent_dist") and sample_mode == "sample": - return encoder_output.latent_dist.sample(generator) - elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax": - return encoder_output.latent_dist.mode() - elif hasattr(encoder_output, "latents"): - return encoder_output.latents - else: - raise AttributeError("Could not access latents of provided encoder_output") - - class QwenImageEditPipeline(DiffusionPipeline, QwenImageEditPipelineMixin, QwenImageLoraLoaderMixin): r""" The Qwen-Image-Edit pipeline for image editing. diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_inpaint.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_inpaint.py index cf802e303e..9e721d0307 100644 --- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_inpaint.py +++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_inpaint.py @@ -12,7 +12,6 @@ # See the License for the specific language governing permissions and # limitations under the License. -import inspect from typing import Any, Callable, Dict, List, Optional, Union import numpy as np @@ -28,7 +27,13 @@ from ...utils import is_torch_xla_available, logging, replace_example_docstring from ...utils.torch_utils import randn_tensor from ..pipeline_utils import DiffusionPipeline from .pipeline_output import QwenImagePipelineOutput -from .pipeline_qwen_utils import QwenImageEditPipelineMixin, calculate_dimensions +from .pipeline_qwen_utils import ( + QwenImageEditPipelineMixin, + calculate_dimensions, + calculate_shift, + retrieve_latents, + retrieve_timesteps, +) if is_torch_xla_available(): @@ -65,94 +70,6 @@ EXAMPLE_DOC_STRING = """ """ -# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift -def calculate_shift( - image_seq_len, - base_seq_len: int = 256, - max_seq_len: int = 4096, - base_shift: float = 0.5, - max_shift: float = 1.15, -): - m = (max_shift - base_shift) / (max_seq_len - base_seq_len) - b = base_shift - m * base_seq_len - mu = image_seq_len * m + b - return mu - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps -def retrieve_timesteps( - scheduler, - num_inference_steps: Optional[int] = None, - device: Optional[Union[str, torch.device]] = None, - timesteps: Optional[List[int]] = None, - sigmas: Optional[List[float]] = None, - **kwargs, -): - r""" - Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles - custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. - - Args: - scheduler (`SchedulerMixin`): - The scheduler to get timesteps from. - num_inference_steps (`int`): - The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps` - must be `None`. - device (`str` or `torch.device`, *optional*): - The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. - timesteps (`List[int]`, *optional*): - Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed, - `num_inference_steps` and `sigmas` must be `None`. - sigmas (`List[float]`, *optional*): - Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed, - `num_inference_steps` and `timesteps` must be `None`. - - Returns: - `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the - second element is the number of inference steps. - """ - if timesteps is not None and sigmas is not None: - raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values") - if timesteps is not None: - accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accepts_timesteps: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" timestep schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - elif sigmas is not None: - accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accept_sigmas: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" sigmas schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - else: - scheduler.set_timesteps(num_inference_steps, device=device, **kwargs) - timesteps = scheduler.timesteps - return timesteps, num_inference_steps - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents -def retrieve_latents( - encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample" -): - if hasattr(encoder_output, "latent_dist") and sample_mode == "sample": - return encoder_output.latent_dist.sample(generator) - elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax": - return encoder_output.latent_dist.mode() - elif hasattr(encoder_output, "latents"): - return encoder_output.latents - else: - raise AttributeError("Could not access latents of provided encoder_output") - - class QwenImageEditInpaintPipeline(DiffusionPipeline, QwenImageEditPipelineMixin, QwenImageLoraLoaderMixin): r""" The Qwen-Image-Edit pipeline for image editing. diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_plus.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_plus.py index f1a069c434..07963c09a3 100644 --- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_plus.py +++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_plus.py @@ -12,7 +12,6 @@ # See the License for the specific language governing permissions and # limitations under the License. -import inspect from typing import Any, Callable, Dict, List, Optional, Union import numpy as np @@ -27,7 +26,13 @@ from ...utils import is_torch_xla_available, logging, replace_example_docstring from ...utils.torch_utils import randn_tensor from ..pipeline_utils import DiffusionPipeline from .pipeline_output import QwenImagePipelineOutput -from .pipeline_qwen_utils import QwenImageEditPlusPipelineMixin, calculate_dimensions +from .pipeline_qwen_utils import ( + QwenImageEditPlusPipelineMixin, + calculate_dimensions, + calculate_shift, + retrieve_latents, + retrieve_timesteps, +) if is_torch_xla_available(): @@ -67,94 +72,6 @@ CONDITION_IMAGE_SIZE = 384 * 384 VAE_IMAGE_SIZE = 1024 * 1024 -# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift -def calculate_shift( - image_seq_len, - base_seq_len: int = 256, - max_seq_len: int = 4096, - base_shift: float = 0.5, - max_shift: float = 1.15, -): - m = (max_shift - base_shift) / (max_seq_len - base_seq_len) - b = base_shift - m * base_seq_len - mu = image_seq_len * m + b - return mu - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps -def retrieve_timesteps( - scheduler, - num_inference_steps: Optional[int] = None, - device: Optional[Union[str, torch.device]] = None, - timesteps: Optional[List[int]] = None, - sigmas: Optional[List[float]] = None, - **kwargs, -): - r""" - Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles - custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. - - Args: - scheduler (`SchedulerMixin`): - The scheduler to get timesteps from. - num_inference_steps (`int`): - The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps` - must be `None`. - device (`str` or `torch.device`, *optional*): - The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. - timesteps (`List[int]`, *optional*): - Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed, - `num_inference_steps` and `sigmas` must be `None`. - sigmas (`List[float]`, *optional*): - Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed, - `num_inference_steps` and `timesteps` must be `None`. - - Returns: - `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the - second element is the number of inference steps. - """ - if timesteps is not None and sigmas is not None: - raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values") - if timesteps is not None: - accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accepts_timesteps: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" timestep schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - elif sigmas is not None: - accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accept_sigmas: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" sigmas schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - else: - scheduler.set_timesteps(num_inference_steps, device=device, **kwargs) - timesteps = scheduler.timesteps - return timesteps, num_inference_steps - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents -def retrieve_latents( - encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample" -): - if hasattr(encoder_output, "latent_dist") and sample_mode == "sample": - return encoder_output.latent_dist.sample(generator) - elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax": - return encoder_output.latent_dist.mode() - elif hasattr(encoder_output, "latents"): - return encoder_output.latents - else: - raise AttributeError("Could not access latents of provided encoder_output") - - class QwenImageEditPlusPipeline(DiffusionPipeline, QwenImageEditPlusPipelineMixin, QwenImageLoraLoaderMixin): r""" The Qwen-Image-Edit pipeline for image editing. diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_img2img.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_img2img.py index f48fbd2a71..fc0b852e83 100644 --- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_img2img.py +++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_img2img.py @@ -1,4 +1,3 @@ -import inspect from typing import Any, Callable, Dict, List, Optional, Union import numpy as np @@ -13,7 +12,7 @@ from ...utils import is_torch_xla_available, logging, replace_example_docstring from ...utils.torch_utils import randn_tensor from ..pipeline_utils import DiffusionPipeline from .pipeline_output import QwenImagePipelineOutput -from .pipeline_qwen_utils import QwenImagePipelineMixin +from .pipeline_qwen_utils import QwenImagePipelineMixin, calculate_shift, retrieve_latents, retrieve_timesteps if is_torch_xla_available(): @@ -44,94 +43,6 @@ EXAMPLE_DOC_STRING = """ """ -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents -def retrieve_latents( - encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample" -): - if hasattr(encoder_output, "latent_dist") and sample_mode == "sample": - return encoder_output.latent_dist.sample(generator) - elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax": - return encoder_output.latent_dist.mode() - elif hasattr(encoder_output, "latents"): - return encoder_output.latents - else: - raise AttributeError("Could not access latents of provided encoder_output") - - -# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift -def calculate_shift( - image_seq_len, - base_seq_len: int = 256, - max_seq_len: int = 4096, - base_shift: float = 0.5, - max_shift: float = 1.15, -): - m = (max_shift - base_shift) / (max_seq_len - base_seq_len) - b = base_shift - m * base_seq_len - mu = image_seq_len * m + b - return mu - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps -def retrieve_timesteps( - scheduler, - num_inference_steps: Optional[int] = None, - device: Optional[Union[str, torch.device]] = None, - timesteps: Optional[List[int]] = None, - sigmas: Optional[List[float]] = None, - **kwargs, -): - r""" - Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles - custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. - - Args: - scheduler (`SchedulerMixin`): - The scheduler to get timesteps from. - num_inference_steps (`int`): - The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps` - must be `None`. - device (`str` or `torch.device`, *optional*): - The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. - timesteps (`List[int]`, *optional*): - Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed, - `num_inference_steps` and `sigmas` must be `None`. - sigmas (`List[float]`, *optional*): - Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed, - `num_inference_steps` and `timesteps` must be `None`. - - Returns: - `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the - second element is the number of inference steps. - """ - if timesteps is not None and sigmas is not None: - raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values") - if timesteps is not None: - accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accepts_timesteps: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" timestep schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - elif sigmas is not None: - accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accept_sigmas: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" sigmas schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - else: - scheduler.set_timesteps(num_inference_steps, device=device, **kwargs) - timesteps = scheduler.timesteps - return timesteps, num_inference_steps - - class QwenImageImg2ImgPipeline(DiffusionPipeline, QwenImagePipelineMixin, QwenImageLoraLoaderMixin): r""" The QwenImage pipeline for text-to-image generation. diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_inpaint.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_inpaint.py index afa27bf01e..db85ae5763 100644 --- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_inpaint.py +++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_inpaint.py @@ -1,4 +1,3 @@ -import inspect from typing import Any, Callable, Dict, List, Optional, Union import numpy as np @@ -14,7 +13,7 @@ from ...utils import is_torch_xla_available, logging, replace_example_docstring from ...utils.torch_utils import randn_tensor from ..pipeline_utils import DiffusionPipeline from .pipeline_output import QwenImagePipelineOutput -from .pipeline_qwen_utils import QwenImagePipelineMixin +from .pipeline_qwen_utils import QwenImagePipelineMixin, calculate_shift, retrieve_latents, retrieve_timesteps if is_torch_xla_available(): @@ -47,94 +46,6 @@ EXAMPLE_DOC_STRING = """ """ -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents -def retrieve_latents( - encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample" -): - if hasattr(encoder_output, "latent_dist") and sample_mode == "sample": - return encoder_output.latent_dist.sample(generator) - elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax": - return encoder_output.latent_dist.mode() - elif hasattr(encoder_output, "latents"): - return encoder_output.latents - else: - raise AttributeError("Could not access latents of provided encoder_output") - - -# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift -def calculate_shift( - image_seq_len, - base_seq_len: int = 256, - max_seq_len: int = 4096, - base_shift: float = 0.5, - max_shift: float = 1.15, -): - m = (max_shift - base_shift) / (max_seq_len - base_seq_len) - b = base_shift - m * base_seq_len - mu = image_seq_len * m + b - return mu - - -# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps -def retrieve_timesteps( - scheduler, - num_inference_steps: Optional[int] = None, - device: Optional[Union[str, torch.device]] = None, - timesteps: Optional[List[int]] = None, - sigmas: Optional[List[float]] = None, - **kwargs, -): - r""" - Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles - custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. - - Args: - scheduler (`SchedulerMixin`): - The scheduler to get timesteps from. - num_inference_steps (`int`): - The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps` - must be `None`. - device (`str` or `torch.device`, *optional*): - The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. - timesteps (`List[int]`, *optional*): - Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed, - `num_inference_steps` and `sigmas` must be `None`. - sigmas (`List[float]`, *optional*): - Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed, - `num_inference_steps` and `timesteps` must be `None`. - - Returns: - `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the - second element is the number of inference steps. - """ - if timesteps is not None and sigmas is not None: - raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values") - if timesteps is not None: - accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accepts_timesteps: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" timestep schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - elif sigmas is not None: - accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) - if not accept_sigmas: - raise ValueError( - f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" - f" sigmas schedules. Please check whether you are using the correct scheduler." - ) - scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs) - timesteps = scheduler.timesteps - num_inference_steps = len(timesteps) - else: - scheduler.set_timesteps(num_inference_steps, device=device, **kwargs) - timesteps = scheduler.timesteps - return timesteps, num_inference_steps - - class QwenImageInpaintPipeline(DiffusionPipeline, QwenImagePipelineMixin, QwenImageLoraLoaderMixin): r""" The QwenImage pipeline for text-to-image generation.